U.S. patent number 8,139,722 [Application Number 12/132,000] was granted by the patent office on 2012-03-20 for system and method for expedited resolution of inquiries for an automated agent.
This patent grant is currently assigned to International Business Machines Corporation. Invention is credited to Subrata K. Das, Ea-Ee Jan.
United States Patent |
8,139,722 |
Das , et al. |
March 20, 2012 |
System and method for expedited resolution of inquiries for an
automated agent
Abstract
An automated agent system and method includes an action
classifier which scores a representation of a portion of a user
inquiry based on a knowledge bank. A comparing device compares a
running score of the representation to a threshold to determine if
enough information has been collected to respond early to the
inquiry. An action mechanism provides an early response to the user
prior to the user completing the inquiry if the threshold has been
exceeded.
Inventors: |
Das; Subrata K. (Granite
Springs, NY), Jan; Ea-Ee (Ardsley, NY) |
Assignee: |
International Business Machines
Corporation (Armonk, NY)
|
Family
ID: |
37108477 |
Appl.
No.: |
12/132,000 |
Filed: |
June 3, 2008 |
Prior Publication Data
|
|
|
|
Document
Identifier |
Publication Date |
|
US 20080226058 A1 |
Sep 18, 2008 |
|
Related U.S. Patent Documents
|
|
|
|
|
|
|
Application
Number |
Filing Date |
Patent Number |
Issue Date |
|
|
11106215 |
Apr 14, 2005 |
|
|
|
|
Current U.S.
Class: |
379/88.01;
704/251; 455/563 |
Current CPC
Class: |
H04M
3/527 (20130101); H04M 2201/40 (20130101) |
Current International
Class: |
H04M
1/64 (20060101) |
Field of
Search: |
;379/265.01-266.1,88.01,88.04 ;455/563 ;704/251 |
References Cited
[Referenced By]
U.S. Patent Documents
Primary Examiner: Matar; Ahmad
Assistant Examiner: Asanbayev; Oleg
Attorney, Agent or Firm: Tutunjian & Bitetto, P.C.
Dougherty; Anne V.
Parent Case Text
RELATED APPLICATION INFORMATION
This application is a Continuation application of U.S. patent
application Ser. No. 11/106,215 filed Apr. 14, 2005 now abandoned,
incorporated herein by reference in its entirety.
Claims
What is claimed is:
1. An automated agent system, comprising: an action classifier,
which scores a representation of a portion of a user inquiry while
a user is speaking based on a knowledge bank; a comparing device
which compares a running score of the scored representation to a
threshold to determine if enough information has been collected to
respond early to the inquiry; an action mechanism which provides an
early response to the user prior to the user completing the user
inquiry by interrupting the user if the threshold has been
exceeded; and an end of speech module configured to detect an end
of speech of the user and generate a final score, and further
configured to compare the final score to a final threshold, such
that if the final score exceeds the final threshold action is made
on the user's inquiry in a no barge-in mode.
2. The system as recited in claim 1, wherein the knowledge bank
includes information based on a history of previous inquiries.
3. The system as recited in claim 1, further comprising a speech
recognizer to generate the representation.
4. The system as recited in claim 1, wherein the action mechanism
includes a barge-in module, which enables the system to barge-in on
the user inquiry prior to its completion.
5. The system as recited in claim 1, further comprising a dialog
manager, which responds to the user's inquiry
6. The system as recited in claim 1, wherein the end of speech
module generates a final score, which is compared to a final
threshold, such that if the final score exceeds the final threshold
action is made on the user's inquiry in a no barge-in mode,
otherwise the user is prompted to repeat the inquiry.
7. The system as recited in claim 1, wherein the running score is
continuously monitored by the comparing device to ensure an
earliest response to the user's inquiry
8. An automated call agent system, comprising: a speech recognizer,
which receives speech information from a user and decodes the
speech information into a representation of the speech information
while the user is speaking; an action classifier, which scores
portions of the representation of the speech information based on a
knowledge bank; a comparing device which compares a running score
of the scored portions of the representation to a threshold to
determine if enough information has been collected to interrupt the
user to respond early to a user inquiry; an action mechanism which
interrupts the user and provides an early response to the user
prior to the user completing the user inquiry, if the threshold has
been exceeded; and an end of speech module configured to detect an
end of speech of the user and generate a final score, and further
configured to compare the final score to a final threshold, such
that if the final score exceeds the final threshold action is made
on the user's inquiry in a no barge-in mode.
9. The system as recited in claim 8, wherein the knowledge bank
includes information based on a history of previous inquiries.
10. The system as recited in claim 8, further comprising a dialog
manager, which responds the user's inquiry when the action
mechanism barges-in.
11. The system as recited in claim 8, wherein the end of speech
module generates a final score, which is compared to a final
threshold, such that if the final score exceeds the final threshold
action is made on the user's inquiry in a no barge-in mode,
otherwise the user is prompted to repeat the inquiry.
12. The system as recited in claim 8, wherein the running score is
continuously monitored by the comparing device to ensure an
earliest response to the user's inquiry.
13. A method for responding early to a user inquiry in an automated
agent system, comprising the steps of: scoring a representation of
a portion of speech information from a user while the user is
speaking based on a knowledge bank using a processor; comparing a
running score of the scored representation to a threshold to
determine if enough information has been collected to respond early
to a user inquiry; and responding to the user prior to the user
completing the user inquiry by interrupting the user; and detecting
an end of speech for the user, including generating a final score,
which is compared to a final threshold, such that if the final
score exceeds the final threshold action is made on the user's
inquiry in a no barge-in mode.
14. The method as recited in claim 13, wherein the knowledge bank
includes information based on a history of previous inquiries, and
further comprising updating the knowledge bank after an action is
taken.
15. The method as recited in claim 13, further comprising
recognizing speech to generate the representation.
16. The method as recited in claim 13, wherein responding includes
barging in on the user inquiry prior to its completion.
17. The method as recited in claim 13, wherein detecting the end of
speech includes generating a final score, which is compared to a
final threshold, such that if the final score exceeds the final
threshold action is made on the user's inquiry in a no barge-in
mode, otherwise the user is prompted to repeat the inquiry.
18. The method as recited in claim 13, wherein comparing includes
continuously monitoring the running score to ensure an earliest
response to the user's inquiry.
19. A program storage device readable by machine, tangibly
embodying a program of instructions executable by the machine to
perform method steps for responding early to a user inquiry in an
automated agent system, as recited in claim 13.
Description
BACKGROUND
1. Technical Field
The present invention relates to an automated agent system and
method and more particularly to a system and method, which
expedites resolution of inquiries in an automated agent.
2. Description of the Related Art
A call center provides rapid resolution of a caller's inquiry. In a
traditional call center manned by human agents, both the caller and
the agent tend to barge in on each other during the course of a
normal conversation; this may serve to speed up the process of
problem resolution.
For example, the following exchange illustrates a call made to a
call center. The caller had just installed a software package on
his computer following the agent's instructions and inquired, "So I
need to start the . . . ", but before the caller could finish the
sentence, the agent barged in with the answer, "Yeah, you need to
take out the CD first and then restart . . . " and so on, since the
agent could guess what the caller was going to say and did not want
to wait for the caller to finish saying the full sentence.
This type of barge-in by the agent and the caller helps to expedite
the process of problem resolution in a traditional call center
manned by human agents.
In an automated call center, barge-in by a caller is often
permitted to speed up the process of problem resolution. For
example, an experienced caller familiar with the call center
functions may decide to bypass the initial greetings playback by
barging-in in the middle of the greetings with an inquiry. However,
the technology for a barge-in function by an automated agent does
not currently exist.
SUMMARY
An automated agent system and method includes an action classifier
which scores a representation of a portion of a user inquiry based
on a knowledge bank. A comparing device compares a running score of
the representation to a threshold to determine if enough
information has been collected to respond early to the inquiry. An
action mechanism provides an early response to the user prior to
the user completing the inquiry if the threshold has been
exceeded.
A method for responding early to a user inquiry in an automated
agent system includes scoring a representation of a portion of a
user inquiry based on a knowledge bank, and comparing a running
score of the representation to a threshold to determine if enough
information has been collected to respond early to the inquiry. A
response is made to the user prior to the user completing the
inquiry.
These and other objects, features and advantages will become
apparent from the following detailed description of illustrative
embodiments thereof, which is to be read in connection with the
accompanying drawings.
BRIEF DESCRIPTION OF DRAWINGS
The disclosure will provide details in the following description of
preferred embodiments with reference to the following figures
wherein:
FIG. 1 is a block diagram of an automated agent that expedites
responding to user inquiries in accordance with one illustrative
embodiment; and
FIG. 2 is a block/flow diagram for an illustrative system/method,
which expedites response time to a user inquiry.
DETAILED DESCRIPTION OF PREFERRED EMBODIMENTS
Illustrative embodiments of the present disclosure provide systems
and methods for barge-in operation by an automated agent to
expedite the problem resolution process. In a particularly useful
embodiment, when a caller initiates an inquiry, the automated agent
keeps recognizing the input speech in conjunction with a knowledge
bank and tries to make a reasonable guess regarding the caller's
intent. Information in the knowledge bank is updated continuously,
either from a call center management team, if they know of
prevalent problems, or it could be culled from caller inputs, if a
number of special issues dominate the calls.
As soon as the automated agent is confident, it barges in on the
caller with a response to the inquiry. Without this type of
barge-in method, the process of problem resolution is likely to be
slower, making the caller unhappy with the automated call center
experience.
The invention described in this disclosure enables barge-in by an
automated agent with the help of a knowledge bank to speed up the
process of caller's problem resolution. One advantage is that the
caller's problem is resolved faster, making the caller satisfied
with the call center experience.
It should be understood that the elements shown in the FIGS. may be
implemented in various forms of hardware, software or combinations
thereof. Preferably, these elements are implemented in a
combination of hardware and software on one or more appropriately
programmed general-purpose digital computers having a processor and
memory and input/output interfaces.
It should be further understood that the use of the word inquiry
includes any uttered speech from a user that can evoke a response
by the agent system described herein. Inquiry may include a user
question, a command, a comment, or a response as the case may
be.
Referring now to the drawings in which like numerals represent the
same or similar elements and initially to FIG. 1, an automated call
center 10 is shown in accordance with an illustrative embodiment.
Speech input 12 from a caller is processed by a speech recognizer
14 and an action classifier 16 to determine an action to be taken.
The action classifier 16 is aided in its task by a knowledge bank
18, which knows about a domain and a state of the inquiry. The
knowledge bank stores information about historically asked
questions or responses and may be trained in accordance with the
specific application(s) to be performed by agent 10.
Action classifier 16 preferably scores using likelihoods or
probabilities that a particular response or inquiry will be posed
to agent 10. This calculation or determination may be performed in
a plurality of ways. For example, a likelihood that a portion of
the response is actually part of a known question is checked and a
score is computed. Known algorithms may be employed. An action
classifier score is output from the classifier 16 and checked
against a threshold in block 20.
In a conventional system this check would be performed only after
the end of the entire inquiry. If the final score falls below a
preselected confidence threshold, the caller is typically prompted
to repeat the query, as the conventional system was unable to make
a proper decision. The caller could then repeat the inquiry or make
a new one as needed and receive a reply from the automated agent
and so on. This suffers from unnecessary delay as described
above.
In accordance with embodiments of the present invention, the speech
input 12, which is processed by the speech recognizer 14 and the
action classifier 16 depends on the knowledge bank 18 to keep track
of the domain and state of the inquiry. When an action is taken,
the knowledge bank is updated to reflect the latest status of a
call and learn new responses.
Rather than waiting until the end of speech is reached, the action
classifier score output from classifier 16 is monitored
continuously in block 20. When the running score in block 20
exceeds a certain confidence threshold, the automated agent 10
permits a barge-in, even though the caller has not finished
speaking the whole inquiry. This includes that when the running
score exceeds the threshold, the automated agent 10 has enough
information to make a decision and respond to the caller.
As long as the running score is less than the threshold, the system
10 keeps checking to see if the caller has finished speaking. When
the end of speech is detected in block 24, the rest of the
procedure may follow the paths as described herein. Thus, the final
score is compared against a confidence threshold in block 26. If
the final score is below this threshold, the caller is prompted to
repeat the query in block 28. Otherwise, action is taken in block
30 in a no barge-in mode.
In block 20, when the running score exceeds the threshold, a
barge-in action is taken in block 32.
The system 10 permits automated agent barge-in, and the action
classifier 16 responds in an appropriate manner in accordance with
the system's understanding of the user's inquiry, expediting the
problem resolution process.
In an illustrative example, a call center receives a call and the
user provides input (12) in the form of an inquiry, for example,
"Is tax included in the price of the item?" The speech recognizer
14 would begin recognizing the speech, and when "Is tax" is
recognized and input to the action classifier 16 using the
knowledge bank 18 would score these two words with a score of say
78%. In this example, the threshold in block 20 is 75%. Therefore,
the running score of 78% exceeds the threshold of 75%.
The action of barging-in would be performed in block 32, followed
by a response generated by a dialog manager 36 or action classifier
16, which generates an appropriate response. For example, in the
present case, the response would be "No, tax is not included in the
price".
Referring to FIG. 2, a block/flow diagram illustratively depicts a
system/method for responding early to a user inquiry in an
automated agent system in accordance with illustrative embodiments.
In block 202, recognizing speech input by a caller in portions as
they are received is optionally performed. Text or other input form
may be provided instead.
In block 203, the speech representation portions from the user
inquiry are scared by the action classifier based on a knowledge
bank. The representation preferably includes recognized or decoded
speech, as it is being received. Faster decoding improves
efficiency of the present invention. The speech is preferably
decoded in portions so that an early inference or guess may be made
as to the inquiry, response or comment being made by the user.
In block 204, a running score of the representation (portions of
the speech) is compared to a threshold to determine if enough
information has been collected to respond early to the inquiry.
This preferably includes periodically or continuously monitoring
the running score to ensure an earliest response to the user's
inquiry.
In block 206, if enough information has been received, prior to the
user completing the inquiry an appropriate response is made. This
may include, e.g., barging in on a user's inquiry prior to its
completion, in block 208.
In block 210, when an end of speech for the user's inquiry is
detected by conventional means, e.g., by monitoring the speech
energy, the final score from the action classifier is compared to a
threshold in block 211. This may occur if an earlier barge-in or
response cannot be made. Detecting the end of speech includes
generating a final action classifier score, which is compared to a
final threshold, such that if the final score exceeds the
threshold, response is made on the inquiry in block 212 in the no
barge-in mode. Otherwise, the user is prompted to repeat the
inquiry or give a new command or statement in block 214.
In block 216, the knowledge bank includes information based on a
history of previous inquiries, and is updated after an action is
taken. Updating the knowledge bank may be employed to increase the
response capabilities of the system.
Having described preferred embodiments of a system and method
system and method for expedited resolution of inquiries for an
automated agent (which are intended to be illustrative and not
limiting), it is noted that modifications and variations can be
made by persons skilled in the art in light of the above teachings.
It is therefore to be understood that changes may be made in the
particular embodiments disclosed which are within the scope and
spirit of the invention as outlined by the appended claims. Having
thus described aspects of the invention, with the details and
particularity required by the patent laws, what is claimed and
desired protected by Letters Patent is set forth in the appended
claims.
* * * * *